This proposal represents an attempt to explain language pathology in dementia in terms of a breakdown in the organization of semantic memory. Demented subjects and aphasics presenting comparable word finding difficulties will be contrasted in a series of experiments intended to reflect upon the state of the semantic representation. We anticipate that on these tasks anomic-aphasics will be indistinguishible from normals, their problem restricted to a lack of access to the internal lexicon. Anomic-demented subjects, on the other hand, are expected to preform significantly worse, but in a manner consistent with models of semantic memory organization. In particular, we predict that semantic memory breaks down in dementia in such a way that concepts become represented in a very vague sense; their relationship to other concepts blurred, their distinguishing characteristics lost.